Bayesian estimation of dynamic panel data gravity model
开发了动态面板数据引力模型的贝叶斯估计方法，处理零贸易流、滞后因变量和多重未观测效应，并重新检验GATT/WTO成员资格对贸易的同期影响，发现加入滞后变量后模型拟合更好，但成员资格无显著同期效应。
In this paper, we develop Bayesian estimation method for inference of dynamic panel data gravity model. Our method deals with the many zeros problem and at the same time, allows for lagged dependent variables and multiple sets of unobserved effects. We apply our Bayesian estimation algorithm to reexamine the contemporaneous effect of GATT/WTO membership on trade. We find that our dynamic gravity model fits the data better than the same model without the lagged dependent variables that is often used in the literature and trade flow in the previous period has a large and positive effect on trade flow in the current period. We also find that the GATT/WTO membership does not appear to have a contemporaneous effect on trade flow. This result is consistent with the findings ofsome studies in the literature, but not with those of others. These results show the importance of including lagged dependent variables and multiple sets of unobserved effects in gravity model estimation.